cross-country studies of growth and policydocuments.worldbank.org/curated/en/... · the recent boom...
TRANSCRIPT
PONCY, R..arh, and Eiea Affairs
WORKING PAPERS
Macroaoonomlo Adjustmontand Growth
Country Economics DepartmentThe World Bank
March 1991WPS 608
Cross-Country Studiesof Growth and Policy
Methodological, Conceptual,and Statistical Problems
Ross Levineand
David Renelt
The design, implementation, and interpretation of cross-countryinvestigations should be improved. This review of conceptual,methodological, and statistical weaknesses in cross-countrystudies suggests that existing findings warrant only limitedconfidence.
Thc Policy, Research. and External Affaus Complex distributes PRE Working Papers to dissaninate the ftndings of work in proges and.o encourage ihe exchangc of ideas among Bank staff and all (vheTs interested in developMent IssuCS These papers carry the names ofthe alhon rsc'd r ait on!)hcsr v ewh and should hc used and cited accordingly The finduing, miterprctations, and conclusions are thea3UhNri' o ri Thc shoild not hc attnhsutod to the World Bank, its loard of D.rectors. its management, oT any of its member countnes
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Polly, Recwoh, and Extel Afbfs
Macroonomlc Adjustmentand Growth
WPS 608
This paper - a product of Fe Macroecoomic Adjusunent and Gowth Clivision, Country EconomicsDeparnment-is pan of a largereffort inPgEto underand empirically the links between policy and long-run growth. Copies are available free from the World Bank, 1818 H Street NW, Washington DC 20433.Please contact CECMG staff, room Ni 1-024, extension 39175 (54 pages).
Levine and Renelt review the conceptual, They hope to stimulate improvements in themethodological, and statistcal problems associ- design, implementation, and interpretability ofated with drawing inferences from cross-country cross-country investigations and to cautionregressions. They elaborate on the particular readers about the confidence they place inproblems associated with empirica ttempts to existing findings.link particular policies with long-run growth.
(See also WPS 609.)
The PRE Working Paper Series disseminates the findings of work under way in the Bank's Policy, Research, and ExtemalAffairsComplex. Anobjective oftheseries is to get these findings outquickly, even if presentations are less than fully polished.The findings, interpretations, and conclusions in thcsc papers do not necessarily represent official Bank policy,
P'roduced by the l'RE Dissemination Center
TABLE OF CONTENTS
I. Introduction 1
II. General Issues 3
A. Aggregation and Sampling 4
B. Interpreting Coefficients 9
C. Causality 10
D. Measurement 11
1. Measurement Error, Index Problems, Data Availability 112. International Price Indexes vs. Own Currency Prices 13
E. Sensitivity Analyses & Fragile Results 18
III. Linking Policy and Growth 20
A. Trade Policy 21
1. Theory 222. Empirical Studies 233. Conclusions on Trade-Growth Literature 28
B. Fiscal Policy 29
C. Financial Policy and Monetary Policy 35
D. Economic Stability, Institutions, and Property Rights 39
IV. Conclusions 41
References 42Table 1 50Table 2 54
Helpful comments were received from Bill Easterly and Lant Pritchett.
I
I. Introduction
The recent boom in theoretical analyses of long run growth has sparked
research and interest in cross-country empirical studies of growth. These
studies typically regress the average rate of growth for a sample of countries
on a group of explanatory variables. The variety of cross-country growth
regressions is enormous. A quick glance at Table 1 indicates that authors
study different sets of countries, over different years, and use different
explanatory variables. The great diversity of studies makes it difficult both
to discern consistent relationships and to compare the results of studies.
Furthermore, the analytical problems that plague cross-country regressions
make it difficult to consider any set of findings reliable.
This paper has two purposes. First, by discussing the methodological,
conceptual, and statistical problems associated with large cross-country
studies of growth, we hope to stimulate improvements in the design,
implementation, and therefore, the interpretability of cross-country
investigations. Although we do not make specific recommendations, we outline
general ways to (a) enhance the econometric design of large cross-country
regression analyses, (b) improve the data construction and the sample
selection processes, and (c) enhance the presentation and interpretation of
cross-country studies. Also, this paper may serve as a useful - though
perhaps overly detailed - warning to readers of the growth literature about
the confidence they should place in cross-country findings.
This paper's second purpose is to examine the particular problems
associated with cross-section attempts to link macroeconomic policies with
growth. Although we would like to construct objective, internationally
2
comparable measures of "fis.wal policy," "trade policy,' "financial policy,"
etc., it is difficult to construct precise empirical measures of these
aggregate concepts for a very broad collection of countries. Consequently,
researchers frequently resort to using measures of economic performance like
the share of exports in GDP. Since these measures of economic performance are
not measures of specific economic policies, cross country growth regressions
involving these performance measures cannot quantify the links between
specific policy changes and growth. Moreover, Levine and Renelt (1990) show
that there is not a strong independent relationship between almost every
existing economic performance measure and long-run growth.
The next section of the critique, Part II, discusses methodological,
conceptual, and statistical problems with cross-country regressions. In
particular, we discuss sampling, aggregation, the interpretation of
coefficients, causality, the selection of data, the implications of
measurement error, and econometric techniques designed to extract particular
components of the data. This entire section may be skipped by readers who are
uninterested in general analytical problems with cross-country regressions.
Part III of this paper provides a review and critique of cross-country
empirical attempt to link policy and growth. We review the theoretical ties
between a host of macroeconomic policies and growth, the empirical findings
regarding each of these policies, and the particular problems associated with
interpreting existing empirical studies of growth and policy.
3
II . General Issues
Many cross-country empirical studies of growth regress the average rate
of growth for a collection of countries on a set of explanatory variables.
While having a common form, empirical studies of growth are not homogeneous:
authors examine different subsets of an enormous number of "right-hand-side"
variables, use different countries, measure variables differently, employ
iifferent data sets, and aggregate data over different periods. Consequently,
it is difficult to discern consistent relationships. Furthermore,
methodological, statistical, and conceptual quandaries aggravate the problem
of drawing reliable inferences from the extensive growth literature.
This section collects, extends, and discusses the analytical.
shortcomings of cross-country studies in the hopes of stimulating improvements
in the design, implementation, and interpretability of cross-country research.
Although cross-country studies may be no place for the methodological
perfectionist [Harberger 1987], authors should be able to answer "yes" to the
following four questions if readers are to have confidence in the results:
1. Are countries the appropriate unit of study?
2. Can the coefficients in a cross-country regression be interpreted in
an economically meaningful way?
3. Are the data used in the regression measured accurately and do they
appropriately represent the concepts for which they proxy?
4. Have extensive sensitivity analyses been conducted to establish the
robustness of the results?
4
This secti discusses each of these questions in some detail. It
should be pointed out that most of the concerns raised in this section have
been discussed by the conductors of cross-country studies themselves. This
section's major practical theme (expressed in Section II.E.) is that
insufficient effort has been devoted to examining tho fragility of existing
empirical conclusions. The econometric work in Levine and Renelt (1990)
suggests that almost all cross-country tegression results are fragile: they
are not robust to slight alterations in the list of explanatory variables.
A. Azgregation and Sampling
In thinking about studying the determinants of growth, a very basic
question comes to mind: What is the appropriate unit of study? In particular,
are countries the appropriate unit to study, or should we conduct analysis at
a more disaggregated level? Since countries are composed of productive
sectors, any country's growth rate depends on world-wide trends in the
country's key sectors as well as country-specific factors that influence all
sectors within the economy.' Decomposing growth into country and sectoral
components would quantify the importance of country-specific factors in
growth.2 Furthermore, empirically isolating the country- and sector-specific
components of grcth is important even if se are only interested in dissecting
the country-specific component of growth. For example, cross-country
'Similarly, the performance of any industry within a given country dependson worldwide trends in that industry and country-specific factors that influenceall industries within that country.
2 Stockman (1988) cor.ducts such a decomposition for seven European countriesand the United States. He finds that there are both important industry-specificand country-specific components to industrial performance. It would beworthwhile to extend this analvsis to more countries.
S
regressions of aggregate growth rates on macroeconomic policy indicators that
do not account for each country's sectoral compositior and world-wide sectoral
trends may capture spurious relationships or miss signi icant relationships
because of the added variability in growth induced by international factors.'
Those attempting to guantify the explanatory power of macroeconomic indicators
associated with fiscal. monetary, trade and financial policy might obtain more
accurate and easily interpretable results by first expunging the comDonent of
national growth reflectinE world-wide trends in the country's major sectors .4
A second problem with using a country as the unit of study involves
sampling. Regression analysis presupposes that the data are samp.ed from a
single "population." It is not clear, however, whether countries are indeed
drawn from the same population. Harberger (1987) asks "hhat do Thailand, the
Dominican Republic, Zimbabwe, Greece, and Bolivia have in common that merits
their being put in the same regression analysis?" Harberger concludes almost
"nothing at all" and warns that "He who puts them in the same regression,
should have a very good reason for doing so." (p.256)
3 For example, consider some macroeconomic policy regime P, and assumethat P positively affects industry 1 and negatively affects industry 2. Alsoconsider two sets of countries all with policy regime P: countries in group Aare composed of industry 1, while group B countries are composed of industry 2.In a cross-country regression with a sample dominated by group A, researchersmight inappropriately conclude that policy regime P should be maintained byeveryone, including group B countries.
Econometrically decomposing national growth rates has its own problems:it may be impossible to obtain sufficiently disaggregated data for manycountries; also, if technology differs significantly across nations, thedecomposition technique employed by Stockman will not isolate the country-specific component of growth [See: Stockman (1985), p. 404-8;. Nonetheless,given the general lack of success in empirically linking macroeconomic indicatorswith average growth rates in cross-sectional regressions [see: Levine and Renelt(1990), extracting the nation-specific component of growth and re-running thecross-sectional analyses with this - albeit problematic - measure of the nation-specific growth rate seems like a worthwhile endeavor.
6
This question of whether countries belong in the same sample is, at
least implicitly, recognized by many researchers when they exclude major oil-
exporting countries, or isolate 'developing" countries, or sort countries by
continent. In excluding major oil-exporting countries, Barro (1989) argues
that 'These countries tend to have high values of -eal GDP per capita, but ...
act more like countries with lower values of income. This behavior can
pLobably be explained by thinking of these countries as receiving large
amounts of income from natural resources, but otherwise not being advanced in
terms of technology, human capital, and so on." (p.18) Alternatively, one may
argue that it is not the large percentage of income received from natural
resources that motivatas the exclusion of major oil-exporting countries from
the sample but the fact that these countries experienced large terms of trade
changes.' Wlile plausible, these arguments imply that countries receiving a
"large" amount of income from natural resources or countries experiencing
"large" terms of trade shocks are not part of the sample of countries we have
in mind in considering a particular set of hypotheses. Should we then exclude
all countries that :eceive some specified amount of income from exporting
natural resources (not just oil producers) and all countries that experienced
sufficiently large terms of trade shocks?
Another example of our difficulty in deciding which countries belong in
the same regression is the frequent separation of developing from
industrialized countries. Table 1 gives a list of cross-coantry studies of
5 An important statistical problem with major oil exporting countries isthat many data sets use 1980 prices. Consequently, when initial income iscomputed in 1950 or 1960 measured initial income for major oil exportingcountries is huge and does not appropriately reflect relative income levels inthese earlier years.
7
growth and indicates whether the samples include only developing countries,
only industrialized countries, or both. Those researchers that distinguish
between developing and industrialized countries are expressing their belief
that countries with different per capita in,mes are sampled from different
populations. It is not clear, however, that we should arbitrarily use per
capita income levels to distribute countries into different 'population'
groups. Furthermore, the delineation between developing and industrialized
countries by per capita income yields troubling categorizations. For example,
this delineation states that Spain and Japan are drawn from the same
population, and Portugal and South Korea are drawn from the same population,
but Spain and Portugal. (as well as Japan and South Korea) are drawn from
different populations.6 Although there is no simple resolution to the
guestion of what countries to include authors should discuss their reasons
for choosing a particular sample and provide information about how the results
change when they use different selection criteria.
Before concluding this discussion of sampling, we raise the problem of
time aggregation: Should a variable measured over long time periods be
aggregated into a single data point? If yes, what should be the frequency of
aggregation? Presumably, the answer d'--ends on what we are trying to measure.
Since most modern theories of growth are discussed in terms of steady-state
solutions, we tend to think about growth as movement along a steady-state
path. Consequently, when we turn to the data, we generally seek to measure
I Grier and Tullock (1989) show that countries, grouped by the OECD and thenby continent, should not be pooled - given the basic macroeconomic variables forwhich they control.
8
"long-run trends" in output as opposed to (a) cyclical variation in production
or (b) transitional movements toward a steady-state
Separating the "trend" component of production from the "cyclical"
component requires that we have a good idea of the frequency of cycles.7 In
practice, taking averages over twenty year periods does not lo gross injustice
to our notions of what is a business cycle and what is not. Howeve-, when we
perform pooled cross-section, time-series investigations using five year
averages, it is less clear that we have eliminated cyclical components.'
A second time-aggregation problem involves distinguishirng steady-state
growth from transitions toward the steady state growth path. Unfortunately,
empirically distinguishing "long-run" trends from transitions toward these
trends is less amenable to reasonable empirical approximations than expunging
cyclical variation from GDP. Our "new" growth models predict that structural
changes such as tax changes, government expenditure changes, and alterations
in the legal system can alter steady-state growth rates. If such changes
occur frequently, the economy will be continually adjusting towards a series
of ever changing steady-state paths. This would imply that we can never
emptrically capture the notion of steady-state GDP movements contained in our
models - and our minds - because we never observe them. Some suggestive
7 This problem is confounded in cross-country studies because countries may
not have the same cyclical frequencies. Furthermore, if we define cycles asperiodic, self-generating movements in out--ut, the profession has not yetconcluded that cycles exist. However, as lot., as the steady-state path is fixed,aggregation in the absence of cycles will still capture steady-state trends.
On the other hand, cycles may be importantly related to steady-stategrowth. Thus, "expunging" the "cyclical" component of output may remove
important information about the long-run growth process.
I This problem is further complicated because business cycle frequenciesmay be different across countries and across variables within the same country.
9
empirical evidence regarding our ability to capture long-run steady-state
trends is provided in Table 2. Using growth rates over five year periods,
Table 2 shows that cross-period correlations are weak and in some cases
negative. Thus, the variables in our regressions may not conform with the
steady-state notions of our theories.
B. InterRreting Coefficients
A second methodological issue that should be confronted when conducting
cross-section regressions is interpreting the coefficients. In many types of
econometric work, coefficients represent our estimates of elasticities or
behavioral relationships. We can then perform conceptual experiments of the
kind: if x changes by one percent our estima as indicate that y will change by
about 'beta' percent. This is not the case with cross-section regressions.
Cross-rountry regressions do not represent behavioral equations; they do not
'describe a single piece of machinery through time." [Harberger 1987, p.256]
Ram (1986) warns that parameter estimates reflect intercountry averages
and do not apply to any single country. In addition, coefficients are not
structural parameters: the sign of an estimated coefficient is "the sign of
the partial correlation between ... [growth] ... and each regressor, with the
other regressors held constant. The corresponding t-statistic ... then can be
taken as a test of the strength of the partial correlations." [Koumendi and
Meguire, 1985, p.146] Consequently, cross-country regressions may best be
viewed as establishing pattertns of correlations. Only theory provides us with
a means of interpreting these patterns. Of course, different theories may
have different explanations for any given set of correlations. By
systematically expanding the set of stylized correlations, however, cross-
10
country empirical studies may be able to favor some existing theories over
others and broaden the equiremeents of future theories.9
C. Causality
The problem of establishing causal relationships in economics is
familiar: Does money cause output, cr does output cause money? Does economic
prosperity foster financial market innovations, or do improvements in
financial arrangements stimulate eco,nomic activity? The list of such
questions is almost endless. Given the quantity of variables used on the
"right-hand-side" of cross-country growth regressions, the problem of
interpreting causal linkages is particularly acute. We agree with Romer's
(1989b) belief that cross-country regressions can only be interpreted within
the context of a theory and that causality only acquires economic content when
we have a theoretical framework for understanding the relationship.'°
While it seems almost self-evident that we need economic theory to
interpret statistical relationships in an economically meaningful way, the
growth literature to date has not optimally integrated econometrics with
economic theory. We have an impressive array of theoretical papers, each
9 Non-parametric techniques can be used to enhance our understanding of
patterns in the data without using regressions. A nice example of this is inDervis and Petri (1987) where they show that " 'high growers' (that is, countriesselected by the criterion of rapid growth) invest like the 'high investors'
(countries selected by the criterion of high investment ratios.)" [Harberger
1987, p.2 56]. Illustrative characterizations of the data help establish
empirical relationships without implying that the results apply to any individual
country or suggesting that the results are structural.
10 Of course, there are empirical definitions of causality. See, forexample, Engle, Hendry, and Richard (1983). But, these definitions also only
acquire economic significance when there is a theoretically meaningful way to
describe the empirically defined causal relationship.
II
motivated by a few stylized correlations; and, we have an impressive array of
cross-country empirical studies, each advertising a few stylized correlations.
wlat we have not had is a sufficiently intensive study of the fragility of
frequently used stylized correlations or a systematic empirical competition
among competing theoretical models of growth.
D. Measurement
This subsection discusses statisti-al and conceptual issues associated
with measuring and constructing the data used in cross-country growth studies.
The subsection is divided into two parts. The first part examines both the
general implications of measurement error and specific biases associated with
cross-country regressions. The second part involves a rather lengthy
discussion of the statistical and conceptual differences between using data
constructed by the International Comparisons Project [see: Summers and Heston
(1988) and Kravis and Lipsey (1990)] or own currency price data available from
the International Monetary Fund and the World Bank. This section may be
skipped by readers not interested in a detailed discussion of the data.
1. Measurement Error. Index Problems, and Data Availability
A troubling problem with cross-country studies of growth is the
measurement of the underlying data. Measurement error generally biases
coefficient estimates. Although data problems plague much empirical work,
measurement problems may be particularly important in cross-country growth
studies because the data collection processes in many countries are poor.
Furthermore, the accuracy of data collection may be correlated with factors
12
such as administrative competence, country size, economic structure, economic
policies (that promote black market activity), and political instability that
may be correlated with economic growth and the level of development.
One example of how measurement error can induce a spurious result
involves the estimation of income in an initial year. Neoclassical growth
theory predicts that income levels of similar countries should converge so
that the coefficient in a cross-country regression of growth on initial income
should be negative. Romer (1989a) points out that when initial income is mis-
measured that the estimated coefficient on initial income in a regression of
average growth on initial income will be biased towards being negative. The
usual approach to errors-in-variables is to instrument the mis-measured
variable. Romer (1989a) does this and finds the negative partial correlation
of initial income with growth disappears when he uses the number of radios and
the consumption of newsprint per capita as instruments. However, these may
not be adequate proxies for initial income and the measurement error in these
variables may be correlated with measurement error in income.
There also exist index number problems in analyzing growth rates as the
structure of production and relative prices may be changing as a country
undergoes growth and structural transformation. For example, the relative
price of capital intensive manufactured goods to labor intensive agricultural
goods or services tends to be greater at lower income levels than higher
income levels. If the manufacturing sector grows more rapidly than the
agricultural or service sectors during economic development, the reported real
rate of growth will be overstated if initial period prices are used and
understated if later period prices are used. A simple chain index can
mitigate this index number problem.
1 3
Another problem is that many policy variables are not directly
measurable. Kormendi and Meguire (1985, p.157) state, "The main issue
requiring more attention is how accurately some of our variables reflect the
hypothesized phenomena." Many institutional aspects of a country such as the
protection of property rights and functioning of the legal system may be
important for growth but cross-country comparative measures of these factors
are weak. This also applies to important policy variables such as the degree
of financial market and trade liberalization. The policy stance of a country
is often inferred from looking at real interest rates, real exchange rates,
effective rates of protection, trade shares, etc., but these measures may not
appropriately reflect underlying policies." By carefully constructing
measures of policy and examining the sensitivity of results to competing
measures of the same policy, researchers could improve our ability to document
the empirical relationships between growth and policy.
2. International Price Indexes vs Own Currency Price Indexes
It is very difficult to compare national incomes. The most common way
to make international comparisons is to convert GDP numbers in own currency
terms into a common currency using exchange rates. This may be
unsatisfactory, however, because nominal exchanges rates generally do not
reflect the real purchasing powers of currencies. For example, if it takes 10
frances to buy in France what $1 will buy in the United States, and the
exchange rate is 5 frances per dollar, then the conversion of French income
expressed in frances into dollars using the exchange rate will overstate
France's real income (from Kravis (1984, p.2]. Furthermore, as Kravis (1984)
"See Pritchett (1990) for a discussion of trade openness measures.
14
and Kravis and Lipsey (1988) discuss, there are good reasons to believe that
own currency prices systematically exaggerate real income differences between
rich and poor countries because non-traded goods are less expensive in poorer
countries but nominal exchange rates do not reflect these differences.
The purpose of the International Comparison Program is eo create
internationally comparable figures for real GDP, the components of real GDP,
and to produce purchasing power parity exchange rates [see: Summers and Heston
(1988) and Kravis and Lipsey (1990)]. Purchasing power parity exchange rates
are the rates that would have to be used such that a ziven basket of goods
that cost $X in the United States (the numeraire country) could be purchased
in any other country - say Chile - by changing the X dollars into Chilean
pesos at the PPP exchange rate and then buying the identical basket of goods
in Chile. Thus, the International Comparisons Project (ICP) collects prices
for about 150 basic categories of goods for a large number of benchmark
countries and then constructs purchasing power parity indexes. These PPP
indexes are aggregated to form PPPs for summary groups of goods (e.g., capital
goods) and further aggregated up to GDP. These PPP indexes are then used to
compare national incomes. Extrapolation is used to provide estimates for
countries not covered by benchmark studies, and, since prices are not
collected every year, ICP extrapolates PPP indexes intertemporally.
Although thes2 data are subject to data limitations, index number
problems, questions associated with the extrapolation techniques, and
quandaries associated with distinguishing the quality of the "same" good in
different countries, Heston and Summers (1988), Kravis and Lipsey (1990), and
Marris (1984) argue that international prices provide a better estimate of
15
income levels for intcrnational comparisons than using exchange rates.12
Fcr example, Kravis and Lipsey (1990, p.4) argue that "the purchasing power of
the currencies of low income countries is much greater than that indicated by
exchange rates ... the real income per capita of the Asian countries was twice
that suggested by exchange rate conversions and that of the Central and South
American countries was half again as much as the exchange rate conversions
indicated." Even so, Kravis and Lipsey (1990) suggest that methodological
problems could cause estimat's to be off by 20-25% for low-income benchmark
countries and 30-35% for non-benchmark countries.
Although ICP prices may provide more easily comparable estimates of
income levels in a given year than one can obtain using exchange rates, it is
much less clear whether ICP prices should be used to compute growth rates or
economic ratios like the share of investment or government spending in GDP.
We first discuss growth rates - following closely the presentation in
Kravis and Lipsey (1990, p.32). The most common way of computing real growth
rates is by deflating each country's nominal GDP figure by its own GDP price
deflator and, then, determining the growth rate of this real GDP statistic.
Since every country's deflator is defined by a different basket of goods, the
growth rate of every country measures the change in a basket of goods that is
different from that measured by all other countries. Thus, this own-currency-
computed growth rate answers the question "How much change has there been in
the quantity of the base year bundle of goods produced in a country?"
t2 The organizers of the ICP, Robert Summers, Robert Lipsey, Irving Kravis,
and Alan Heston, are very careful to document the procedures they use in
constructing indexes, and they highlight any problems about which they think
users of the data should be aware.
16
Computing growth rates using international prices answers a different
question. For ICP-computed growth rates, the questions is "What is the change
in the value of the goods produced in a country where value is computed using
the same international price index for all countries?" "Such growth rates
have the merit of treating a given increase in a given good as making the same
contribution to growth in both countries. They have the drawback that the
prices used may be very dissimilar from the prices of one or both of the
situations." (Kravis and Lipsey, 1990, p.32)
Thus, we tend to agree with Kravis and Lipsey (1990) that ICP growth
rates may better capture notions of international opportunity costs while own
currency growth rates better capture welfare considerations because own
currency prices more accurately reflect the representative consumption baskets
of consumers. Statistically, we found that over the period 1960 to 1985 the
two growth rates have a simple correlation coefficient of almost 0.9.
Turning to economic ratios, there exist important conceptual and
statistical differences in the use of international vs. own currency prices in
computing the shares of various economic variables in GDP. Investment goods
tend to be relatively more expensive than consumption or government services
in low-income countries, while government services are generally more
expensive in high-income countries. This implies that the actual quantity of
investment goods purchased will be smaller in a low-income than a high-income
country for any given proportion of income spent on investment. Thus,
investment measured using international prices may more accurately measure the
augmentation of the capital stock. One problem with using ICP data in
computing iwvestment shares, however, is that international differences in
these prices may also reflect trade and fiscal policies which may
17
inappropriately affect the pricing of capital goods.'' De Long and Summers
(1990) point out that some of the newly industrializing countries such as
Korea and Brazil have relatively low prices of investment durables relative to
construction prices. They argue that this is a result of different trade and
tax policies in these countries designed to promote industrialization. The
effects of policy on investment prices may be particularly important for
countries for which benchmark studies of prices have not been done as one may
be estimating effects based on extrapolating from other countries policies.
Even for benchmark countries this could be important as extrapolations from
the benchmark year to other years may not fully account for policy changes.
The issue of choosing ICP or own country data may be even more
problematic in the case of government expenditures because there are various
interpretations of what the government expenditure to GDP ratio is supposed to
measure. Ram (1987) argues that ICP prices are better than own currency
prices when one is trying to measure government provision of productive
services because own currency prices will not account for the tendency of
government services to be relatively inexpensive in low-income countries.
Thus, the ratio of government expenditures to GDP computed using own currency
prices may under-estimate the provision of public goods in low-income
countries. On the other hand, if one is trying to measure the distortionary
effects of taxation, one may prefer own country prices because these may more
accurately reflect the relative size of t:he government in the economy.
Statistically, the simple correlations of growth and government
consumption share over the period 1960-85 are strongly negatively correlated
using the data constructed by Summers-Heston and positively correlated using
'3See Bradford (1987), Barro (1989), and De Long and Summers (1990).
18
World Bank data. Because of the conceptual and empirical differences
associated with using either own currency prices (associated ith World Bank
and 1MF data) or ICP prices (associated with the Summers and Heston data set),
researchers should make specific arguments for using one set of data rather
than the other and may even want to compare results using both sets of data.
E. Sensitivity Analysis & Fragile Results
Doing sensitivity analysis means addressing the question: "Do the
conclusions withstand slight alterations in the right-hand-side variables, in
functional form, serial correlation assumptions, measurement error processes,
distributional assumptions, sample period, and the weighting of observations?
We must subiect our econometric studies to systematic sensitivity analyses to
determine whether the results are fragile or robust: i.e.. sensitivity
analyses helD determine the extent to which we believe econometric studies.
While pioneers in the field of sensitivity analyses such as Edward
Leamer may complain about the haphazard way we study the fragility of
empirical results, many cross-country empirical studies do not suffer from
this criticism because they have failed to analyze the robustness of their
results, haphazardly or otherwise. For example, B.rro (1989, 1990, 1991), Ram
(1986), and Landau (1983, 1986) focus on the relationship between government
expenditures and growth, but they never test whether their findings would
change if they included proxies for tax policy, trade policy, or financial
policy. Similarly, Tyler (1981), Feder (1983), and Moschos (1989) study the
growth effects of trade policy, but they do not examine the sensitivity of the
results to the inclusion of variables that represent fiscal, monetary, or
exchange rate policies. A quick glance at Table 1 indicates that most studies
19
focus on a limited set of relationships and do not inquire whether the results
are sensitive to other explanatory variables.
The lack of sensitivity analysus is particularly troublesome because of
the serious methodological, conceptual, and statistical problems with cross-
country regressions discussed above. We need to investigate extensively the
sensitivity of our findings to slight changes in the list of explanatory
variables, small reweighting of observations, minor alterations in assumptions
about the distribution of residuals, different measurement error processes,
etcetera.14 Levine and Renelt (1990) present evidence showing that slight
alterations in the list of explanatory variables can over turn the results
found in many empirical growth studies. This finding surely tempers the
confidence we should place in the conclusions of existing studies.
14 See Leamer (1983, 1987) for a general discussion of sensitivity analysesand for citations that exemplify different types of sensitivity analyses.
20
III. Linking Policy and Growth
This section discusses the problems associated with empirically linking
policy with growth. Researchers have used measures of fiscal, monetary,
trade, financial, and exchange rate policies as well as indicators of the
institutional, legal, and political character of countries in cross-countty
regressions. This section reviews the theoretical ties between each policy
and growth. We then critique existing empirical findings and summarize their
methodological, conceptual and statistical problems.
Although each subsection focuses on particular problems associated with
linking specific policies with gtowth, some general themes emerge from the
discussion. First, it is very difficult to construct objective, continuous,
interaationally comparable measures of macroeconomic policies such as "fiscal
policy" or "trade policy" or to construct sharp empirical measures of concepts
such as "the efficiency of the legal system" or "the degree of political
stability." Second, since it is difficult to measure policy directly,
researchers resort to measures of performance such as the share of exports in
GDP or the share of government consumption in GDP. Therefore, we must be
careful not to interpret tiiese measures as indicators of specific policies and
we must be particularly careful in drawing causal inferences. For erample,
the common finding that the growth of exports is positively related to the
growth of output does not necessarily indicate that export promotion policies
stimulate growth because (1) we have not related any particular export
promotion policy to growth; and (2) all findings using exports can be
replicated using imports or total trade [Levine and Renelt (1990)]. Finally,
almost none of the findings in cross-country studies of growth is robust to
21
slight alterations in the list of explanatory variables. Considerably more
cross-section work needs to be done to confidently link policy with growth.
Detailed case studies may be able to more accurately capture salient policy
differences than broad cross-country studies.
A. Trade Poligy
This section summarizes the conceptual and statistical difficulties
associated with cross-country studies that attempt to link trade policy and
growth. Four important themes run though this review. First, theory
typically analyzes the relationship between trade and growth, but empirical
work frequently focuses on the relationship between exports and growth.
Econometrically, this does not induce important problems because exports and
imports are highly correlated, so that all of the results obtained using
exports can be obtained using imports or total trade. Using exports instead
of trade becomes a problem when authors interpret their results as
establishing an exclusive empirical relationship between exports and growth
instead of a general relationship between trade and growth. Second, policy
makers are concerned about the relationship between trade policy and growth,
but many empirical studies do not examine trade policy; they examine the
correlation between exports and growth. We must, therefore, be very careful
not to interpret these studies as quantifying the effects of trade policy
(particularly export promotion policies). Third, it is very difficult to
quantify trade policy with objective, continuous, internationally comparable
proxies. Consequently, although studies that measure trade policy address an
important policy question, the measurement problems are so severe [see:
Edwards (1989) and Pritchett (1990)] that a skeptical reader could justifiably
22
conclude that the links between trade policy and growth are tenuous. Finally,
the conclusions of many growth studies can be easily overturned by slightly
altering the list of explanatory variables.
1. Theory
Economists' concern with the relationship between international trade
and economic growth extends back at least until Adam Smith. Smith focused on
the gains in productivity that result from increases in specialization, and
noted that openness to international markets could encourage specialization.
International trade may enhance productivity by allowing agents to specialize
in activities that would be unprofitable in smaller markets and by allowing
countries to specialize in fields in which they have comparative advantages.
Along these lines, Rivera-Batiz and Romer (1989) and Grossman and Helpman
(1989) have recently constructed rigorous models in which technology is
produced in profit maximizing firms. They show that openness to international
markets can increase the growth rate of technology by increasing the size of
the market available to technology producers and allowing those countries with
a comparative advantage in technology production to specialize in this key
industry. Romer (1986, 1990) also notes that international trade ray improve
domestic productivity and economic growth by increasing communication with and
therefore "knowledge spillovers" from trading partners.
Focussing on the negative effects of trade distortions, Krueger (1974)
Grossman and Helpman (1989, NBER 2970), Murphy, Shleifer, and Vishny (1990)
show that quotas may divert talented people out of productive activities and
into rent-seeking endeavors. This distortion can slow the rate of
tachnological improvement and retard growth. Similarly, Corden (1974)
discusses the conditions under which trade restrictions may induce
23
entrepreneurs in protected sectors to alter their labor-leisure choices and
work less. Although in most models trade distortions retard economic growth,
Grossman and Helpman (1989) point out circumstances under which certain trade
distortions accelerate growth.'" Thus, there is an important empirical
question regarding the relationship between various trade policies and
economic growth.
2. EmRirical Studies
In a thorough review of the trade-growth literature, Sebastian Edwards
(1989) divides the empirical growth literature on trade policy into three
categories:1" papers that attempt to link measures of either export growth or
the role of exports in the economy to growth; investigations that use
theoretical models to compute continuous, objective measures of trade
distortions and then link these measures of trade intervention with growth;
and large multicountry case-studies that attempt to link indicators of trade
orientation with growth.
In synthesis volumes of multicountry-case studies, Krueger (1978),
Bhagwati (1978), Balassa (1982), and The World Bank (1987) carefully document
the experiences of many countries that have undergone trade li.beralization
efforts. Although a wealth of information is contained in these studies,
Edwards (1989) argues that methodological and statistical problems dampen the
'"For example, in a world where technology is produced and where technologyhas external effects, a country with a comparative advantage in technologyproduction can increase world growth by protecting the technology sector.
la Edwards (1989) also considers studies that try to uncover thedeterminants of exports.
24
confidence we should attach to their conclusions regarding the favorable
relationship between trade liberalization and growth. Nonetheless, a careful
reading of these studies would make it very difficult for even a skeptical
reader to conclude that trade liberalization does not hold favorable
implications for growth.
The second group of empirical studies identified by Edwards (1989) are
those that attempt to link measuro.s of trade policy with growth. Papers by
Krueger (1983), Havrylyshyn (1985), and Edwards (1989) compute the
discrepancies between observed trade and the predictions of the Heckscher-
Ohlin model and use these discrepancies as indicators of trade policy.
Although this procedure produces continuous, objective, and internationally
comparable indicators of trade policy, these measures depend on the adequacy
of the Hechscher-Ohlin model. Furthermore, the resulting estimates of trade
intervention display troublesome patterns. For example, the intervention
measure used by Edwards (1989) is Leamer's (]c.87) intervention index. This
measure of intervention, however, is significantly positively correlated with
Leamer's measure of openness. Using a different approach, Dollar (1990)
constructs an index of the distortion between domestic and international
prices and shows that this index is correlated with growth in a cross-section
of countries. In a thorough study of trade policy indicators, Pritchett
(1990) concludes that there do not exist reliable cross-country estimates of
trade policy. Similarly, Pack (1988) and Rodrik (1989) conclude that there is
no clear cut confirmation that countries open to international trade enjoy
more rapid productivity growth."7
"7 Also see Levine and Renelt (1990).
25
The third and most heavily populated group of empirical studies
identified by Edwards (1989) are papers that focus on the relationship between
exports and growth. This group can be subdivided into (1) studies that use
the share of the export sector in domestic output (or the change in the share
of the export sector in output) [Romer (1989, 1990) and Kormendi and McGuire
(1985)] and (2) studies that use measures of export growth [Balassa (1978),
Tyler (1981), Feder (1983), Kavoussi (1984), Ram (1985), and Moschos (1989)].
Romer (1990) and Kormendi and McGuire (1985) - who use measures of the
share of exports in GDP - explicitly state that they are attempting to proxy
for the importance of international activity in the country's economic life,
i.e., they are trying to measure openness, not exports per se. Thus, we are
supposed to interpret the coefficients in these regressions as addressing the
question: "Do countries with relatively more economic interactions with the
world community grow faster." Or if one is using the growth rate in the share
of exports in GDP, the question becomes "Do countries where the proportion of
economic activity conducted with the rest of the world increases quickly enjoy
faster growth rates than other countries?" We must, however, recognize that
(1) differences in the fraction (or differences in the growth rate of the
fraction) of exports to GDP may not be tied to policy changes, and (2) the
ratio of exports or trade to CDP may not be a good indicator of growth
inducing economic interactions with the international community.
Empirically, there is a fairly robust two-step empirical link between
the share of exports in GDP and output growth. For a very diverse set of
specifications, Romer (1990b) and Le-vine and Renelt (1990) demonstrate that
the share of exports in GDP is significantly positively correlated with the
ratio of investment expenditures to GDP and that the investment ratio is
26
significantly positively correlated with per capita output growth."8 These
findings should be interpreted as establishing a robust two-step partial
correlation between the trade share and growth because the results using trade
share (or import share) are equal to those obtained using exports. This
partial correlation between trade share and growth, of course, does not tell
us much about trade policy, and, in particular, these findings do not tell us
anything about export promotion policies.
The second category of growth studies that focus on the relationship
between exports and growth use measures of export growth in their growth
regressions. They typically find a positive coefficient. Theoretical
justification for using export growth (or the interaction term export growth
times the share of exports in GDP) is provided by Feder (1983). He assumes
that there are positive externalities associated with exports. Thus, the
production of non-export goods is assumed to depend positively on the
production of exports. Many studies show that output growth is positively
correlated with export growth. While there may inideed be externalities
associated with exports, some readers may be bothered by a regression that
includes the growth rate of two variables in the national accounts identity.
Since the basic components of the GNP accounts will probably covary positively
over long time periods, regressing output growth on export growth may not tell
us much about the importance of international economic relationships in
economic development. Interestingly, Ram (1986) argues that government
provision of public goods provide external benefits to domestic production and
'a The export shares is not significantly correlated with per capita growthwhen the investment share is included in the regression, but the investment shareremains significantly correlated with growth when the export share is included.
27
that governmernt expenditures - like exports - should be included as separate
entries in donestic production functions.'9 Levine and Renelt (1990) combine
the Feder (1983) study of export growth and output growth with the Ram (1986)
study of government spending growth and output growth and show that export
growth enters insignificantly once government expenditure growth and the
growth in government's share of GDP are included.
Turning from empirical to conceptual issues, causality is particularly
problematic in cross-country regressions of output growth on export growth.
For example, a shift in government expenditures from defense to education
might stimulate long-run domestic production, including the production of
exports. Attributing the growth in output to growth in exports would be
inappropriate. Indeed, Jung and Marshall (1985) show that increases (or
decreases) in the growth rate of exports are very poor predictors of increases
(or decreases) in the growth rate of output.
Another weakness with cross-country growth regressions that focus on the
relationship between exports and growth is that they generally do not examine
specific proxies for trade policy and yet they tend to draw conclusions
concerning trade policy in general and export promotion policies in
particular.' Since (1) these studies do not include proxies for trade
policy, (2) the causal relationship between export growth and output growth is
ambiguous, (3) all the empirical relationships obtained by these studies using
'9 Similarly, it is not difficult to envisage the argument that imports arean important source of knowledge spillovers and that imports too belong asseparate entries in the production functions of domestically produced goods.
' Balassa (1985) uses a measure of import penetration as a policyindicator. Pritchett (1990), however, demonstrates the problem with using importpenetration as a measure of trade policy.
28
export growth can be obtained using import growth or trade growth [Levine and
Renelt (1990)], arid (4) the empirical results obtained by these cross-country
studies of growth and export growth break-down when government spending growth
is included [Levine and Renelt (1990)), we should not base our support or
opposition to export promotion policies on existing cross-country growth
regressions involving exports.
3. Conclusions concerning Trade-Crow.h Literature
In summing-up this discussion of cross-country empirical studies of
growth and trade policy three points standout. First, it is difficult to read
the carefully documented muilticountry studies by Krueger, Bhagwati, and
Belassa and believe that trade policy and growth are unrelated. Yet, the
conceptual and statistical problems with these studies discussed by Edwards
(1989) makes one reluctant to conclude that "trade liberalization promotes
growth." Second, although attempts at constructing objective, continuous,
internationally comparable proxies of trade policy have thus far been plagued
by crippling measurement problems, the potential benefits of constructing
"good" proxies for trade policy is enormous. It is only with "good" proxies
for trade policy that we can address the relevant question: what is the
relationship between specific trade policies and growth. Finally, studies of
the relationship between exports and growth suffer from a number of
interpretational problems: (1) one can substitute imports or trade everywhere
in these papers without changing the results; (2) these studies have no direct
link to trade policy in general or to export promotion policies in particular;
(3) the conclusions of these studies are typically not robust to the inclusion
29
of other policy variables; and (4) the causal ties between exports and output
arc ambiguous both theoretically and empirically.
B. Fiscal Policy
One of the most important issues in economics is the role of government
expenditures and taxation in economic growth. A large number of cross-
country studies of growth have attempted to link aggregate measures of fiscal
policy with average annual growth rates computed over long time periods [see:
Table 1]. Thus far, the literature has been generally unsuccessful in
identifying robust empirica_ relationships between growth and aggregate
indicators of government expenditures or taxes. This section briefly reviews
the theoretical relationships between fiscal policy and growth, the conceptual
complexities associated with using macroeconomic theory to guide cross-
country empirical investigations of growth and fiscal policy, and some
statistical reasons why we have been unsuccessful in identifying consistent
empirical relationships between existing measures of fiscal policy and growth.
There are three basic themes in this review. First, the intuition underlying
the theoretical linkages between fiscal policy and growth is intuitively
appealing and fairly straightforward. Second, it is difficult to measure
government provision of services because aggregate measures of expenditures do
not delineate among categories of government expenditures that may have very
different growth implications, and money spent may not accurately represent
the actual delivery of services. Third, a country's "tax system" is difficult
to represent using aggregate measures because the structure of taxes,
enforcement, and the tax base differ internationally. Evaluating fiscal
30
policy is complex and not easily amenable to broad, internationally comparable
macroeconomic indicators.
The recent endogenous growth literature has created a new class of
models in which fiscal policy can have long-run steady-state growth effects.
The ideas behind these models, however, h 'e been around for a very long time.
Governments can accelerate growth by providing essential public goods, and
well-designed taxes and subsidies can close the gap between private and social
costs. On the other hand, government funds may be spent on activities for
which there is not a clear role for the government. Thus, broad measures of
government expenditures may not appropriately measure the provision of growth-
inducing social services. Furthermore, even if one obtains more disaggregated
data on government expenditures, funds may be spent effectively or
ineffectively, so that using simple expenditure data without accounting for
government efficiency may be very misleading. Similarly, it is very difficult
to construct meaningful measures of something as complex as a country's "tax
system" while appropriately considering international differences in the
structure of taxes and the size of the tax base.
Even putting aside the differential growth effects of different types of
government expenditures and the differential growth effects of different types
of taxes, there may be complex tradeoffs between the beneficial effects of
government services and the deleterious effects of distortionary taxes. In
Barro (1990) and Easterly (1990), growth increases with taxation and
expenditures at low levels and then decrease as the distortionary effects of
taxation exceed the beneficial effects of public goods. Government
expenditures and growth are positively correlated when government expenditures
are below the optimum amount, negatively related when they are above, and
31
there is no cross-section correlation when governments are providing the
optimal amcunt of services. Unfortunately, cross-country empirical studies
have not exploited this potentially non-linear relationship and instead
estimate simple linear equations.
The above discussion indicates that, even though the intuition
underlying the models of fiscal policy and growth is simple, it is very
difficult to construct informative empirical proxies for social services.
Empirically, Ram (1986) finds a positive correlation between the growth rate
of government expenditures and output growth. It is not clear, however, if
this correlation has much economic content. For example, if the demand for
government services increases with income, one could find a positive
correlation between government expenditures and growth even if greater
government expenditures hamper growth.2" Furthermore, one might find a
positive relationship between the growth rate of government expenditures and
output growth even if the role of government in the economy falls as the
country develops. These interpretational problems have led many researchers
to use more disaggregated measures of government expenditures.
Barro (1989, 1990, 1991) and Diamond (1989) use detailed measures of
government expenditures on capital goods, education, defense, and consumption
spending less defense and education payments. Barro (1991) finds that the
ratio of government consumption expenditures less defense and education
expenditures to GDP is negatively correlated with growth. Levine and Renelt
(1990), however, show that this negative correlation Lzcomes insignificant for
2'A number of papers [see Ram(1987)] have considered Wagner's Law (the share
of government expenditures increases with income) but this need not hold for a
strong positive correlation between output growth and government growth to hold
because if the government share remains unchanged there will be proportionality
between the two growth rates.
32
some econometric specifications. Barro (1991) also finds that the coefficient
on the ratio of government capital expenditure or elucation expenditures to
growth depend on the specific econometric specification employed. Diamond
(1989) tests for separate effects of a number of categories of public
expenditure (over a short period-1980-85) with mixed results. He finds
generally positive effects for capital expenditure in the social and education
sectors (which may augment human capital). The coefficients for other
categories of government expenditures are quite fragile to the inclusion of
other explanatory variables.'
Attempts to capture the effects of taxes on growth have also produced
mixed results. Trying to get an aggregate measure of the potentially negative
implications of government activity, many researchers use government
consumption spending as a proxy for the distortionary taxes that must be
raised to support that spending.2 3 In this case, total government
expenditures and not government consumption expenditures should be used, but
these data are not available for many countries over very long time periods.24
Kormendi and MeGuire (1985) find that the average growth rate of the ratio of
government consumption spending to GDP is not closely associated with growth
and Levine and Renelt (1990) find that both the ratio of government
consumption expenditures and total government expenditures to GDP are not
' This raises further aggregation issues: government transfers, interest
payments, expenditures by different levels of government, and public enterprises
expenses may have different effects from other consumption expenditures.
23See Landau (1983,1986), Romer (1989a,b), and Easterly and Wetzel (1989).
24 Total government expenditures is a concept that varies across countries
depending on whether it includes all levels of government, just the central
government, various categories of state enterprises, etc.
33
strorngly related to growth. These results suggest that more detailed
information on taxes may be necessary to assess the growth effects of taxes.
There are, of course, considerable conceptual and statistical problems
inherent in constructing cross-country measures of taxes. First, it is
difficult to define the tax base. Poorer countries tend to have a smaller
taxable sector which is taxed at higher rates. When GDP is used in the
denominator of tax ratios, it yields a low "average" tax rate for some
countries without capturing the distributional aspects of tax policy. Second,
it is difficult to proxy for the "tax system". There are tremendous
differences in the types of taxes used and the enforcement of the tax code.
Also the structure of taxation tends to vary systematically with income.
Poorer countries rely more on trade taxes and less on personal income taxes.
To the extent that different taxes have different growth effects, using
aggregate measures that miss these differences will not capture salient
aspects of the tax system.' Finally, it is difficult to measure marginal
taxes. This problem is compounded when one considers th - axes differ across
sectors, so that computing a country's marginal tax means somehow aggregating
sectoral taxes into an average national marginal tax rate.
Given these difficulties, some efforts have been made to test for the
impact taxes on growth. For example, Koester and Kormendi (1989) try to
examine the differential effects of marginal and average taxes. They use the
tax to GDP ratio as a measure of average taxes and interpret the regression
coefficient of GDP on taxes as a marginal tax rate. They find that taxes do
not have growth effects. In a second example, Skinner (1987) and Manas-Anton
'See Shah and Whalley (1990) for a discussion of tax structure differencesin developing countries.
34
(1987) analyze the differential effects of direct vs. ind.rect taxation. The
null hypothesis is that indirect taxes tend to fall more heavily on
consumption and therefore have smaller growth effects. Skinner (1987) finds
evidence that individual and corporate taxes have greater negative growth
effects than trade or sales taxes in Africa, but Manas-Anton (1987) finds
little support for a greater negative impact of direct vs. indirect taxes.'
It would appear that tax policy is an area in which more detailed studies will
be needed to evaluate the growth effects of fiscal policy.
The empirical work on fiscal policy and economic growth has not produced
robust empirical relationships. This may be because governments are providing
the optimal amount of public goods. However, the inability of researchers to
establish empirical links between fiscal policy and growth may stem from
inadequate measures of the delivery of public goods or our failure to capture
the relevant characteristics of national tax systems. A detailed analysis of
the composition of government expenditures and the structure of the tax system
may be necessary if we are to link fiscal policy with growth. Unfortunately,
the data needed for such analysis is not readily available for a wide cross-
section of countries.
X Statistically, the severity of direct taxes may be overstated in somedeveloping countries if they fall heavily on economic rents, such as mineralextraction or monopolies.
35
C. Firnancial Paolicy & Monetary Pol icy
Financial irnstruxnents anid financial institut:ions have been integral
parts of economic activity for over two hundred years. The most importarnt
financial instruments have probably been nonmetallic money and demand
deposits, but bonds, equities, options, and forward contracts also play
important economic roles. Similarly, banks have been the most ubiquitous
financial institutions, but mutual funds, investment banks, and insurance
companies are becoming increasingly important in many countries. To the
extent that an economy's financial structure - its composition of financial
instruments and institutions - affects economic growth, financial and monetary
policy may affect growth by altering the financial structure. Theoretically,
however, the profession is only beginning to integrate financial markets into
modern growth theories. And, econometrically, it is very difficult to obtain
good measures of financial policy - or even financial market performance - for
a broad cross-section of countries to include in cross-country growth studies.
The role of money in economic activity is one of the most frequently
studied issues in economics [see: the extensive survey by Orphanides and Solow
in the Handbook of Monetary Economics). Early rational expectations models by
Lucas (1972, 1973) and Barro (1976, 1980) predicted a neutral relationship
between anticipated money growth and output. But, in recent models, high
money-growth, high inflation environments can elicit behaviors that reduce
growth. For example, talented agents may transfer out of productive
enterprises and into rent-seeking activities, agents may substitute out of
simple money exchange and into transactions technologies that require more
time and effort, or capital accumulatinn may be discouraged [Stockman (1981)].
Consequentl;. rapid money-growth could retard growth. On the other hand, the
36
Tobin-Mundell school argues that inflation will induce a substitution out of
money into capital so that growth may increase.
The cross-country empirical evidence on the relationship between money,
growth, and inflation is ambiguous. Kormendi and Meguire concluce that
average growth rate of inflation is negatively correlated with growth, and Ml
growth has little relation to growth. But, Grier and Tullock (1989) find that
both the sign and the significance of the inflation-growth correlation depends
importantly on the sample chosen. Levine and Renelt (1990) demonstrate that
the relationship between growth and inflation and between growth and domestic
credit growth depend on the inclusion of other policy variables.
Although theory suggests that monetary policy variability should impede
the efficient allocation of resources, the empirical relationship between
monetary policy uncertainty and growth is ambiguous. For example, Hayek
(1944), Friedman (1977), and Barro (1976, 198C) argue that variable inflation
or monetary policy uncertainty can interfere with the ability of agents to
extract informa-.)n from relative prices and may reduce investment and
economic performance. Empirically, Kormendi and Meguire (1985) find a
negative correlation between the standard deviation of Ml growth and output
growth. Grier and Tullock (1989), however, again find that both the sign and
significance of this correlation depend on the sample of countries chosen, and
Levine and Renelt (1990) demonstrate that small changes in the explanatory
variables can change the sign of the coefficient on the standard deviation of
inflation or the standard deviation of the growth rate of domestic credit in
cross-country growth regressions. A country's monetary policy may be so
clo-ely linked to the broad macroeconomic policy configuration of the country
37
that simple lin3ar growth regressions can not isolate the effects of monetary
policy on growth.
Outside of money, the fin..ncial system may provide a varietv of
important services. As discussed in the World Development Report 1989,
financial systems can help pool, allocate, and manage risk, mobilize savings
for projects that could not be independently financed, monitor the activities
of mangers, provide information on the economy, and evaluate enterprises for
investors. Although these products of the financial system are well known, it
is only recently that economists have begun to study the emergence and
implications of financial assets and institutions within the context tormal
theoretical frameworks. McKinnon (1973) and Shaw (1973) provide insightful
descriptions of th.e role of finance in growth. Bernanke and Gertler (1989,
1990) formally model the role of financial markets in depressions. Jovanovic
and Greenwood (1990), Greenwald and Stiglitz (1989), and Levine (1990, 1991)
formally model potential linkages between economic growth and financial
markets that emerge to mitigate specific problems that exist in the model
economy, e.g., allocation of risk under asymmetric information, economies of
scale in information gathering, etc.
Econometrically, it is difficult to construct measures of specific
financial policies or of the services provided by financial markets that are
meaningful and available for a broad cross-section of countries. Gelb (1989),
The World Development Report (1989), Goldsmith (1969), and McKinnon (1973)
have linked measures of domestic financial activity to grcwth. For example,
Goldsmith (1969) finds that as real income riser, the ratio of financial
institutions' assets to GNP grows, while The World Development Report 1989
demonstrates that the distribution of financial assets among financial
38
institutions systematically differs when countries are grouped according the
real per capita income, Gelb (1989) presents cross-country regressions
results indicating a positive relationship between the ratio of M3 to GDP for
a sample of 31 countries, and also shows that real interest rates are
positively related to growth - potentially capturing the negative growth
implications of repressed nominal interest rates.
Two problems with these empirical studies - as all of the authors note -
is disentangling causal directions and isolating specific policies that
underlie the performance criteria used in the studies. For example, per
capita output growth and technological change may elicit the creation and
modification of financial arrangements; thus, a positive relationship between
an indicator of financial market activity and growth may not imply that
financial markets cause growth. Similarly, finding a positive correlation
between the M3-GDP ratio and output growth has little to say about specific
financial market policies.
The empirical evidence suggests a positive relationship between
financial market activity and growth, but the profession has been unable to
determine the direction of causality or link specific financial market
policies with economic performance. Given the potential importance of
monetary and financial policy in economic growth, considerably more work
should be devoted to constructing indicators o financial policy and
attempting to link these indicators to growth. Along these lines, one
unexploited and potentially profitable research endeavor would be to
incorporate measures of policies toward international capital - instead of
simply focussing on domestic financial markets - into our growth models and
empirical designs.
39
D. Economic Stability. Institutions and Propertv Rights
Economists have become increasingly concerned about the role of property
rights, economic stability, and political instability in growth. Although
some serious attempts have been made to include measures of these concepts in
cross-country regressions, it is clear that internationally comparable,
objective measures of notions such as "the efficiency of tne legal system" or
"the degree of political uncertainty" may be impossible to obtain.
Nonetheless, we should not ignore these concepts if we are to explain cross-
country differences in growth rates.
The theoretical links between growth and economic stability, political
security, and well-designed, efficiently enforced property rights is well
grounded even though not much formal work has been done within the context of
growth models. Macroeconomic instability, political volatility, and insecure
or poorly enforced property rights create uncertainty as to the ability of
individuals to reap the benefits of investments in physical and human capital.
Thus, increased uncertainty associated with these phenomena will tend to lower
investment and growth. Furthermore, political and legal institut.;ons define
the "rules of the game." If these "rules" are uncertain or severely
burdensome to economic players, it makes the types of complex exchanges that
occur in modern economies very difficult.
Empirically, however, it is very difficult to isolate the effects of
"politics" or "the efficiency of the legal system" on economic activity.27
27 The statistical measurement of instability is also problematic. One yearof severe inflation, for example, in an otherwise stable country may ,.ead to ahigher measure of instability than a country which has numerous bcuts of moderateinflation. Statistically isolating these differences is a hurdle that growtheconomists must face. Similarlv. since data is usuallv measured over long time
40
For example, Londregan and Poole (1990) find that the propensity of coups is
positively related to the level of poverty and also that past coups contribute
to slower growth. Similarly, Barro (1991) finds that social indicators of
war, revolution and civil liberties [also see: Scully (1988) and Kormendi and
Meguire (1985)] are negatively related to growth. But, there is a significant
correlation between these measures of political instability and measure of
monetary or exchange rate variability (about 0.4 with each). This highlights
the difficulties in drawing causal links. It may be that politically unstable
countries produce policy instability or that policy instability leads to
political instability or both. The difficulties inherent in disentangling
these effects and relating them to growth are as difficult as they are
important."
periods, taking averages of measures like revolutions or coups may not adequatelydistinguish continual political unrest from a brief period of intense unrest.
'Haggard, Kaufmann, Shariff, and Webb (1990) discuss economic and politicalinstability links.
41
IV, Conclusion
After reviewing the analytical problems associated with drawing
inferences from cross-country regressions, thi. paper discussed the particular
problems associated with interpreting empirical attempts to link macroeconomic
policy indicators with growth. By collecting at.4 studying the problems with
existing empirical studies of growth, we hope to (1) stimulate researchers to
advance the study of long-run growth, and (2) caution readers about the
confidence they should place in existing findings. Regionally based pooled
cross-section, time-series analyses that allow for more country specific
information while maintaining the ability to use standard econometric
techniques along with detailed longitudinal case studies that rely more on
qualitative analyses may - in conjunction with improved cross-sectional work
- augment our understanding of the linkages between policy and long-run
growth.
42
References
Agarwala, R. 1983. "Price Distortions and Growth in Developing Countries." WorldBank Staff WP #575.
Aghion, Philippe and Peter Howitt. 1989. "A Model of Growth through CreativeDestruction." MIT.
Arrau, Patricio. 1989. "Human Capital and Endogenous Growth in a Large ScaleLife-Cycle Model." World Bank PPR WP #342.
Arrow, Kenneth. 1962. "The Economic Implications of Learning by Doing." Review ofEconomic Studies 29 (June) p. 661-669.
Azariadis, Costas and Allen Drazen. 1990. "Threshold Externalities in EconomicDevelopment." Ouarterly Journal of Economics. (June)
Backus, David, Patrick Kehoe, and Timothy Kehoe. 1990. "In Search of ScaleEffects in Trade and Growth."
Balassa, Bela. 1978. "Exports and Economic Growth: Further Evidence." Journalof Development Economics. 5 (June) p.1 81-18 9.
-..... 1982. DeveloRment Strategies in Semi-Industrial Economies. Baltimore:Johns Hopkins University Press.
------. 1985. "Exports, Policy Choices, and Economic Growth in DevelopingCountries After the 1973 Oil Shock." Journal of Development Economics.18 p.2 3-3 5.
Barro, Robert J. 1991. "Economic Growth in a Cross Section of Countries."forthcoming Ouarterly Journal of Economics.
------. 1990. "Government Spending in a Simple Model of Endogenous Growth."Journal of Political Economy 98, 103-125.
------. 1989. "A Cross-Country Study of Growth, Saving and Government."NBER WP#2855.
-----. 1980. "A Capital Market in an Equilibrium Business Cycle Model."Econometrica. 48. p. 1393-1417.
-..... 1976. "Rational Expectations and the Role of Monetary Policy." Journal ofMonetary Economics. 2. p. 1-32.
-----and Xavier Sala-i-Martin. 1990. "Public Finance in Models of EconomicGrowth." NBER WP #3362.
Becker, Gary, Kevin Murphy, and Robert Tamura. 1990. "Economic Growth, HumanCapital and Population Growth." Journal of Political Economy 98, 12-37.
Bhagwati, J. 1978. Anatomy and Consequences of Exchange Control Regimes.
43
Cambridge MA: Dallinger Publishing Company.
Bradford, Colin Jr. 1987. "NlCs and next-tier NICs as Transitional Economies." inC. Bradford and W. Branson, eds., Trade arid STructural Change in Pacific
Asia. Chicago: NBER Press,
Cardoso, Eliana, and Albert Fishlow. 1989. "Latin American Economic Development:
1950-80." NBER WP#3161.
Chenery, Hollis. 1986. "Growth and Transformation." Ch.2 in H.Chenery, S.
Robinson, and M. Syrquin Industrialization and Growth. World Bank.
De Long, J. Bradford. 1988. "Productivity Growth, Convergence, andWelfare:Comment." American Economic Review December.
----- and Lawrence Summers. 1990. "Equipment Investment, Relative Prices andEconomic Growth." Harvard University.
De Melo, Jaime and Sherman Robinson. 1989. "Productivity and Externalities:Models of Export-Led Growth." World Bank.
Dervis, Kermal and Peter Petri. 1987. "The Macroeconomics of SuccessfulDevelopment: What are the Lessons?" NBER Macroeconon,ics Annual. p.211-54.
Diamond, J. 1989. "Government Expenditure and Economic Growth: An EmpiricalInvestigation." IMF WP#89/45.
Dollar, David. 1990. "Outward-Oriented Developing Economies Really Do Grow MoreRapidly: Evidence from 95 LDCs, 1976-85." World Bank.
Easterly, William R. 1990a. "How Does Growth Begin? Models of EndogenousDevelopment." World Bank.
.----. 1990b. "Endogenous Growth in Developing Countries with Government InducedDistortions." World Bank.
----- and Deborah L. Wetzel. 1989. "Policy Determinants of Growth: Survey ofTheory and Evidence." World Bank PPR WP#343.
Edwards, Sebastian. 1989. "Openness, Outward Orientation, Trade Liberalization,and Economic Performance in Developing Countries." World Bank PPR WP#l91.
Engle, R.F., D.F. Hendry, and J.-F. Richard. 1983. "ExogEtieity." Econometrica.51. p. 277-304.
Feder, Gershon. 1983. "On Exports and Economic Growth." Journal of DeveloDmentEconomics, 12 p.59-7 4.
Friedman, Milton. 1977. "Inflation and Unemployment." Journal of PoliticalEconomy. 85. p. 451-472.
14 4
Gastil, R.D. 1978. Freedom in the World: Political r.ights and Civil Liberties.New York: Freedom liouse.
Gelb, Alan. 1989. "A Cross-Section Analysis of Financial Policies, Efficiencyand Growth." World Bank.
Goldsmith, Raymond W., 1969. Financial Structure and Development. New Haven: YaleUniversity Press.
Greenwald, B.C. and J.E. Stiglitz. 1989. "Financial Market Imperfections andProductivity Growth." NBER Working Paper No. 2945.
Greenwood, J. and B. Jovanovic. 1990. "Financial Development, Growth, and theDistribution of Income." Journal of Political Economy 98, 1076-1107.
Grier, Keviin and Gordon Tullock. 1989. "An Empirical Analysis of Cross-NationalEconomic Growth, 1951-1980." Journal of Monetary Economics 24 p.2 59-2 76.
Grossman, Gene and Elahan Helpman. 1989a. "Endogenous Product Cycles." NBERWP#2913.
-and -----. 1989b. "Growth and Welfare in a Small Open Economy." NBERWP#2970.
----- and -----. 1989c. "Quality Ladders and Product Cycles." NBER WP#3201.
----- and -----. 1990. "Cc.nparative Advantage and Long-Run Growth." AmericanEconomic Review.
Gupta, Kanhaya and M. Anisul Islam. 1983. Foreign Capital. Savings and Growth: AnInternational Cross-Section Study. Dordrecht,Holland: D.Reidel.
Haggard, Stephan, Robert Kaufman, Karim Shariff, and Steven Webb. 1990. "ThePolitical Economy of Inflation and Stabilization in Middle-IncomeCountries." World Bank.
Harberger, Arnold, 1987. "Comment" on Dervis and Petri (1987).
Havrylyshyn, 0 . 1985. "The Direction of Developing Country Trade: EmpiricalEvidence." Journal of Development Economics. 19, p.25 5-28 1.
Hayek, F.A. 1944. The Road to Serfdom. Chicago: University of Chicago Press.
Heston, Alan and Robert Summers. 1988 "What We Have Learned about Prices andQuantities from International Comparisons: 1987." American Economic Review.78.2 (May) p. 467-473.
Hicks, Norman. 1980. "Economic Growth and Human Resources." World Bank WP#408.
Hwa, Erh-Cheng. 1983. "The Contribution of Agriculture to Economic Growth:SomeEmpirical Evidence." World Bank Staff WP #619.
45
Jung, Woo and P. Marshall. 1985. "Exports, Growth and Causality in Developing
Countries." Journal of Development Economics, 18, p.1-12.
Kavousi, R. 1984. "Export Expansion and Economic Growth: Further Evidence."
Journal of Development Economics. 14. p. 241-250.
Khan, Moshin and Carmen Reinhart. 1990. "Private Investment and Economic Growth
in Developing Countries." World Development 18 p.19-27.
King, Robert and Sergio Rebelo. 1989. "Transitional Dynamics and Economic Growth
in the Neoclassical Model." NBER Working Paper, No. 3185.
Koester, Richard and Roger Kormendi. 1989. "Taxation, Aggregate Activity and
Economic Growth: Cross-Country Evidence on Some Supply-Side Hypotheses."
Economic Inquiry 27 (July) p.3 6 7-8 6.
Kormendi, Roger and Philip Meguire. 1985. "Macroeconomic Determinants of Growth:
Cross-Country Evidence." Journal of Monetary Economics 16 p.14 1-1 6 3.
-----, .---- , and Victor Lavy. 1988. "Aid, Investment, and Economic Growth in
Developing Countries." Hebrew Univ. WP#184.
Kravis, Irving B. 1984. "Comparative Studies of National Incomes and Prices."
Journal of Economic Literature. 22 (March).
------, and Robert Lipsey. 1988. "National Price Levels and the Price of
Tradables and Nontradables." American Economic Review. 78. p.474-478
, and ------- 1990. "The International Comparison Program:
Current Status and Problems." NBER WP#3304.
Krueger, Anne. 1978. Foreign Trade Regimes and Economic Development:Liberalization Attempts and Consequences. Cambridge MA: NBER.
Landau, Daniel. 1983. "Government Expenditure and Economic Growth: A Cross-
Country Study." Southern Economic Journal (January) p.783-92.
-----. 1986. "Goverment and Economic Growth in the Less Developed Countries: An
Empirical Study for 1960-80." Economic Development and Cultural Change 35
(October) p.3 5-7 5.
Leamer, Edward E. 1978. Specification Searches: Ad Hoc Inference from
Non-experimental Data. New York: Wiley.
-----. 1983. "Let's Take the Con out of Econometrics." American Economic Review.
73 (March) p.31-43.
-----. 1985. "Sensitivity Analysis Would Help." American Economic Review. 75
(June) p.308 -113.
-----. 1987. "Errors in Variables in Linear Systems." Econometrica. 55 (July)
p.89 3-909.
46
----- and Herman Leonard. 1983. "Reporting the Fragility of Regression
Estimates." Review of Economics and Statistics. 65 (May) p.306-317.
Levine, Ross. 1991. "Stock Markets, Growth and Policy." forthcoming Journal of
Finance.
-, 1990. "Financial Structure and Economic Development." Board of Governorsof the Federal Reserve System, Washington, D.C.. IFDP #381.
----- and David Renelt. 1990. "A Sensitivity Analysis of Cross-Country Growth
Regressions.' World Bank Mimeo.
Lavy, Victor. 1988. "Aid and Growth in Sub-Saharan Africa: The Recent
Experience." European Economic Review 32 p.1 77 7-17 9 5.
Londregan, John and Keith Poole. 1990. "Poverty, the Coup Trap, and the Seizureof Executive Power." World Politics 42 (January) p.15 1-1 8 3.
Lucas, Robert. 1988. "On the Mechanics of Economic Development." Journal of
Monetary Economics 22 (July) p.3-4 2.
Manas-Anton, L.A. 1987. "Relationship Between Income Tax Ratios and Growth Rates
in Developing Countries: A Cross-Country Analysis." in V. Ghandi ed.Supply-Side Tax Policy. Washington DC: IMF.
Mankiw, N. Gregory, David Romer, and David Weil. 1990. "A Contribution to the
Empirics of Economic Growth." Harvard University.
Marris, Robin. 1984. "Comparing the Income of Nations: A Critique of the
International Comparison P'roject." Journal of Economic Literature. 22
(March) p. 40-57.
Marsden, Keith. 1983. "Links Between Taxes and Economic Growth: Some Empirical
Evidence." World Bank WP#605.
Martin, Ricardo and Mohsen Fardmanesh. 1990. "Fiscal Variables and Growth: A
Cross-Section Analysis." Public Choice 64 (March) p.2 39-2 51.
McKinnon, Ronald I., 1973. Money and Capital in Economic Development. Washington:
Brookings Institution.
Moschos, Demetrios. 1989. "Export Expansion, Growth and the Level of Economic
Development: An Empirical Analysis." Journal of Development Economics 30
p.93-102.
Murphy, Kevin, Adnrei Shleifer, and Robert Vishny. 1990. "The Allocation of
Talent: Implications for Growth." University of Chicago.
Nishimizu, Mieko and John Page Jr. 1990. "Trade Policy, Market Orientation, and
Productivity Change in Industry." in j. de Melo and A. Sapir eds. Trade
Theory and Economic Reform North. South. and East.
47
North, Douglass. 1987. "Institutions, Transaction Costs and Economic Growth."Economic Inquiry. 25 (July) p.4 19-4 2 8.
Nuxoll, Daniel. 1990. "Neo-classical Growth Models and Accelerating GrowthRates." Brown University.
Otani, I. and D. Villanueva. 1990. "Long-Term Growth in Developing Countries andits Determinants: An Empirical Analysis." World DeveloRment 18.
Pindyck, Robert. 1989. "Irreversibility, Uncertainty, and Investment." WorldBank PPR WPS#294.
Pritchett, Lant. 1990. "Measuring Trade Barriers in LDCs: Can It Be Done." WorldBank.
Ram, Rati. 1985. "Exports and Economic Growth: Some Additional Evidence."Economic DeveloRment and Cultural Change 12 p.59-74.
------. 1986. "Government Size and Economic Growth: A New Framework and SomeEvidence from Cross-Section and Time Series Data." American Economic Review76 (March) p.19 1-2 0 3.
-----. 1987. "Wagner's Hypothesis in Time-Series and Cross-Section PerspectivesEvidence from "Real" Data for 115 Countries." Review of Economics andStatistics (May) p.194-204.
Rebelo, Sergio, 1991. "Long Run Policy Analysis and Long Run Growth." forthcomingJournal of Political Economy.
------. 1990. "Growth in Open Economies." unpublished manuscript. The World Bank.
Rittenberg, Libby. 1989. "On the Problem of Identifying the Engines of EconomicGrowth." Studies in Comparative International Development 24 (Fall).
Rivera-Batiz, Luis. 1989. "Growth, Technology, and Trade between AsymmetricCountries." University of Chicago.
----- and Paul Romer. 1989. "International Trade with Endogenous TechnologicalChange." University of Chicago.
Robinson, Sherman. 1971. "Sources of Growth in Less Developed Countries: A Cross-Section Study." Ouarterly Journal of Economics 85 (August).
Rodrik, Dani. 1988. "Closing the Productivity Gap: Does Trade LiberalizationReally Help?" Harvard University.
Romer, Paul. 1986. "Increasing Returns and Long-Run Growth." Journal of PoliticalEconomy. 94 p.10 0 2-1 0 3 7.
------. 1987. "Crazy Explanations for the Productivity Slowdown." in S. Fischered. NBER Macroeconomics Annual Cambridge MA: MIT Press.
48
----. 1989a. "Human Capital and Growth: Theory and Evidence." NBER WP#3173.
-----. 1989b. "What Determines the Rate of Growth of Technological Change?" WorldBank PPR WP#279.
----. 1989c. "Capital Accumulation in the Theory of Long-Run Growth." in R.Barroed. Modern Business Cycle Theory. Cambridge MA: Harvard University.
-----. 1990a. "Endogenous Technological Change.' Journal of Political Economy 98,71-103.
------. 1990b. "Capital, Labor, and Productivity." Brookings Papers on EconomicActivity: Microeconomics.
Scholling, Eberhard and Vincenz Timmermann. 1988. "Why LDC Growth Rates Differ:Measuring "Unmeasurable" Influences." World Development 16 p.127 1-12 9 4.
Schultz, T. 1961. ""Investment in Human Capital." American Economic Review. 51,p.1-17.
Scully, Gerald. 1988. "The Institutional Framework and Economic Development."Journal of Political Economy 96 (3).
Shah, Anwar and John Whalley. 1990. "An Alternative View of Tax IncidenceAnalysis for Developing Countries." World Bank.
Shaw, Edward S. 1973. Financial Deepening in Economic Development. New York:Oxford University Press.
Skinner, Jonathan. 1987. "Taxation and Output Growth: Evidence from AfricanCountries." NBER WP#2335.
-----. 1989. "Do Taxes Matter?: A Review of the Effect of Taxation on EconomicBehavior and Output." World Bank W7S#48.
Solow, Robert. 1956. "A Contribution to the Theory of Economic Growth." OuarterlyJournal of Economics 70 (February).
------. 1957. "Technical Change and the Aggregate Production Function." Review ofEconomics and Statistics 39 p.3 12-3 2 0.
Stockman, Alan. 1988. "Sectoral and National Aggregate Disturbances to IndustrialOutput in Seven European Countries." Journal of Monetary Economics. 21p.387-409.
------. 1981. "Anticipated Inflation and the Capital Stock in a Cash-in-AdvanceEconomy." Journal of Monetary Economics. 8. p. 387-393.
Summers, Robert, and Alan Heston. 1988. "A New Set of International Comparisonsof Real Product and Price Levels Estimates for 130 Countries, 1950-1985."Review of Income and Wealth 34 (March) p.1-2 5.
149
Tyler, L. 1985. "Growth arnd Export Expansion in Developin g Countries.' Journalof Development Economics 9 p.121-30.
W;eede, Erich. 1983. "The Impact of Democracy on Economic Growth: Some Evidencefrom Cross-National Analysis." Kyklos.
Vheeler, David. 1980. "Human Resource Development and Econornic Growth inDeveloping Countries: A Simultaneous Model." World Bank WF#407.
World Bank. 1987. World Development Report. London: Oxford University Press.
World Bank. 1989. World Development Report. London: Oxford University Press.
'rable 1
Cross-Section Studies of Economic Growth
Independent Variables'Study Period #C D.V. IS LG HK IY XG XS CC CK TX FL IN PI PF OV
Balassa (85) 1973-89 43d GY + 0 - + Y
Barro (89a) 1960-85 72 GYP + - 0 - - +/0 - +/0 Y
Barro (89b) 1960-85 94 GYP + + - y
Cardoso & 1950-80P 18d GY + + + N
Fishlow (89)
De Long (88) 1870-79 22 GYP Y
De Long & 1960-85 42 GYP +/0 0 -/0 YSummers (90)
Diamond (89) 1980-85 38d GY + 0 +/0 + 0 0 +/0 N
Dollar (90) 1976-85 95d GYP + y
Easterly & 1960-85 70d GY + +/0 + - + Y
Wetzel (89)
Edwards (89) 1960-82 28d GY + + 0 Y
Feder (83) 1964-73 31d GY + + + N
Gelb (89) 1965-85 34d GY + y
Grier & 1950-81P 24D GYP + - - - Y
Tullock (89) 1960-81P 89d GYP + + - + y
Gupta & 1965-73 52d GY + + y
Islam (83)
Hicks (80) 1960-77 65d GYP + + + y
+ (-) iadicates found significantly positive (negative), 0 indicates insignificant, +/0indicates significant in some regressions, blank indicates variable not included in study
Period: Time period of cross section analyzed, P indicates panel used#C: Number of countries, d indicates limited to developing countries,D to developedD.V.: GY-Growth of GDP, GYP-Growth of per capita GDPI.V.: IS-Investment share of GDP, LG-Labor growth, HK-Human capital variable, IY-Initialperiod income, XG-Growth of exports, XS-! port share, GC-Goverrunent consumption share,GK-Government capital share, TX-Tax variable, FL-Financial liberalization, IN-Inflationvariable, PI-Political instability, PF-Political freedom, OV-Other variables used (Y/N)these are reviewed on the following page. Variable ̂ ontent and definitions may vary acrossstudies.
51
Table 1: Cross-Section S.udies of Economic Growth ContinuedIndependent Variables
St.udv Period #C D.V. IS LG ED IY XG XS GC GK TX FL IN PI PF OV
Hwa (83) 1970-79 87 GY + + + - Y
Khan & 1970-79 24d CY + +/0 + 0 N
Reinhart (90)
Koester & 1970-79 63 GY + + - 0 Y
Kormendi (89)
Kormendi,Lavl., 1968-81 62d GY 0 0 - + y& Meguire (88)
Kormendi & 1950-77 47 GY + + - +/00 - + Y
Meguire (85)
Landau (83) 1961-76 96 GYP + Y
Landau (86) 1960-80P 65d GYP + 0 + - - 0 y
Lavy (88) 1968-82 22d GY + +/0 0 Y
Londregan & 1950-82P 121 GY - -/0 YPoole (90)
Manas-Anton(86)1973-82 39d GY + 0 + 0 0 -/0 - N
Mankiw,Romer 1960-85 98 GYP + 0 + - N
& Weal (90)
Marsden (83) 1970-79 20d GY + + N
Martin & 1972-81 76 GY + + + - Y
Farmanesh (90)
Moschos (89) 1970-80 71d GY + 0 + N
Murphy,Shleiferl970-85 91 GYP + +/0 - -/0 - Y& Vishny (90)
+ (-) indicates found significantly positive (negative), 3 indicates insignificant, +/0
indicates significant in some regressions, blank indicates variable not included in study
Period: Time period of cross section analyzed, P indicates panel used#C: Number of countries, d indicates limited to developing countries,D to developedD.V.: GY-Growth of GDP, GYP-Growth of per capita GDPI.V.: IS-Investment share of GDP, LG-Labor growth, HK-Human capital variable, IY-Initialperiod income, XG-Growth of exports, XS-Export share, GC-Government consumption share,GK-Government capital share, TX-Tax variable, FL-Financial liberalization, IN-Inflationvariable, PI-Political instability, PF-Political freedom, OV-Other variables used (Y/N)these are reviewed on the following page. Variable content and definitions may vary acrossstudies.
TalId C ir,)tS S jitl S dd j & H ( of Economlic r ,rowah ContinuedIlndecpendufi Vartiabl s'
Sucdl Pe r i o( *C D V IS l.G F.D IY XG XS ;C G' CCIX Fl, IN PI PF OV
i & i9 83 ,ti-(t CI I- ) + i
Villanueva (90)
Ram (86) 1960-80 115 GY + 4 +/0 N
Rittenberger(89) 70-82 57d GY + +/0 + Y
Robinson (71) 1958-66 39d GY + 0 Y
Romer (89a) 1960-85 94 GYP + +/0 -/0 - y
Romer (89b) 1960-85 90 GYP + +/C + - Y
Scully (88) 1960-80 115 GYP + N
Skinner (87) 1965-82 29d GY 0 0 - + - - y
Tyler (81) 1960-77 41d GY + + + N
Weede (83) 1960-79 94 GYP + 4 0 Y
Theejer d80) 1960-77 43d GY 4 + 4 y
+ (-) indicates found significantly positive (negative), 0 indicates insignificant, +/0indicates significant in some regressions, blank indicates variable not included in study
Period: Time period of cross section analyzed, P indicates panel used1,C: Number of countries, d indicates limited to developing countries,D to developedD.V.: GY-Growth of GDP, GYP-Growth of per capita GDPI.V.: IS-Investment share of GDP, I.G=Labor growth, HK=Hnman c3pital variable, IY-Initialperiod income, XG=Growth of exports, XS-Export share, GC-Government consumption share,GK=Gover!nnent capital share, TX=Tax variable, FL-Financial liberalization, IN-Inflationvariable, PI-Political instability, PF-Political freedom, OV-Other variables used (Y/N)these are reviewed on the following page. Variable content .nd definitions may vary acrossstudies.
Table 1: Other Variables Included and Results
Balassa (85) Outward Orientatiorl (+), Manuf. Share Exports (t)
Barro (89a,b) Socialist economy (-/'O), Mixed economv ,-/0)Invest. Price deviation (-), Africa (-)I .atin America (-)
De Long (88) Protestant religion (+)
De Long & Summers (90) Investment durables price and share (+)
Dollar (90) Real exchange rate distortion (-) & variability (-)
Easterly & Wetzel (89) Inward trade orientation (-), Africa (-), Latin America (-)
Edwards (89) Trade intervention (-)Gelb (89) Distortion index lAgarwala,19831 (-)Grier & Tullock (89) Variation in oucput growth (+)
Gupta & Islam (83) Foreign Aid(+/O),Foreign Investment(O), Other Foreign Capital (+)
Hicks (80) Life expectancy (+)
Hwa (83) Agriculture growth (+)
Koester & Kormendi (89) Marginal tax (-/0)
Kormendi, Lavy Money growth (0), Variation in output (0), Foreign aid (+/0)
& Meguire (88)Kormendi & Meguire (85) Variation in output (+)
Landau (83) Climate dummies (+/'0)
Landau (86) Population (-),Transfers from abroad (+), Distance to seaport(-)
Lavy (88) Terms of trade (-/0)
Londregan & Poole (90) Africa (-), Europe & North America (+)
Martin & Farmanesh (90) Government deficit (-)Murphy, Shleifer, Engineering students (+/O),Law students (-/0)
& Vishny (90)Otani & Villanueva (90) Interest rate on external debt (0)
Rittenberger (89) Agriculture Growth (+/O), Manufacturing growth (+/O)
Services growth (+)
Robinson (71) Net foreign balances (+), Change in agriculture share (+)
Change in city share of population (+)
Romer (89a,b) Africa (-), Latin America (-)
Skinner (87) Terms of trade (+), Oil (+/O)
Weede (83) Political democracv (-/0), Military (+)
Wheeler (80) Change in nutrition (4)
54
TABLE 2
SUBPERIOD CORRELATIONS OF GROWTH OF REAL PER CAPITA GDP 1960-85
ALL COUNTRIES1l60-85 1960-65 1965-70 1970-75 1975-80 1980-85
1960-85 1 .47 .57 .65 .66 .641960-65 1 .01 .20 .03 .171965-70 1 .29 .37 .141970-75 1 .23 .211975-80 1 .35
NON-OIL COUNTRIES1960-85 1960-65 1965-70 1970-75 1975-80 1980-85
1960-85 1 .52 .60 .64 .74 .691960-65 1 .10 .10 .23 .251965-70 1 .35 .33 .231970-75 1 .32 .251975-80 1 .46
LATIN AMERICA1960-85 1960-65 1965-70 1970-75 1975-80 1980-85
1960-85 1 .15 .68 .63 .57 .751960-65 1 .23 -. 25 -. 36 .341965-70 1 .31 .17 .451970-75 1 .17 .231975-80 1 .28
AFRICA1960-85 1960-65 1965-70 1970-75 1975-80 1980-85
1960-85 1 .50 .53 .68 .62 .621960-65 1 .00 .14 .05 .161965-70 1 .34 .20 .111970-75 1 .30 .221975-80 1 .35
OTHER NON-OIL COUNTRIES1960-85 1960-65 1965-70 1970-75 1975-80 1980-85
1960-85 1 .43 .74 .57 .75 .54l6O-65 1 .17 -.03 .33 -.061965-70 1 .40 .35 .241970-75 1 .17 .051975-80 1 .49
Source: Summers-Heston (1988)
PRE Working Pager Series
Contactijw Author DAO for pape
WPS596 The Mexican Sugar Industry: Brent Borrell February 1991 P. KokilaProblems and Prospects 33716
WPS597 Rent Sharing in the Mufi-Fibre Refik Erzan February 1991 G. lbgonArrangement: Theory and Evidence Kala Krishna 33732from U.S. Apparel Imports from Ling Hlu TanHong Kong
WPS598 Africa Region Population Projections: Patience W. Stephens February 1991 0. Nadora1990-91 Edition Eduard Bos 31091
My T. VuRodoHfo A. Bulatao
WPS599 Asia Region Population Projetions: Eduard Bos February 1991 0. Nadora1990-91 Edition Patience W. Stephens 31091
My T. VuRodoifo A. Bulatao
WPS600 Latin America and the Caribbean My T. Vu February 1991 0. NadoraRegion Population Projections: Eduard Bos 310911990-91 Edition Patience W. Stephens
Rodolfo A. Bulatao
WPS601 Eu.,ope, Middle East, and North Eduard Bos February 1991 0. NadoraAfrica Region Population Projections: Patience W. Stephens 310911990-91 Edition My T. Vu
Rodolfo A. Bulatao
WPS602 Firm Output Adjustment to Trade Mark A. Dutz February 1991 S. FallonLiberalization: Theory with 38009Application to the MoroccanExperience
WPS603 The Role of Officially Supported Asli Demirguc-Kunt February 1991 G. lbogonExport Credis in Sub-Saharan Refik Erzan 33732Africa's External Financing
WPS604 Foreign Trade and Is Relation to Faezeh Foroutan February 1991 S. FallonCompetition and Productivity in 37942Turkish Industry
WPS605 Overview of Contractual Savings Dimitri Vittas March 1991 W. Pitayato-Savings Institutions Michael Skully nakarn
37666
WPS606 Adjustment Policies and Investment Luis Serven March 1991 E. KhinePerformance in Developing Andres Solimano 39361Countries: Theory, Country Experiences,and Policy Implications
PRE Working Pager Series
ContactTg ALLhor DaL for paper
WPS607 Abolishing Green Rates: The Effects Donald F. Larson March 1991 D. Gustafsonon Cereals, Sugar, and Oilseeds Simon Glance 33714in West Germany Brent Borrell
Merlinda IngcoJonathan Coleman
WPS608 Cross-Country Studies of Growth Ross Levine March 1991 CECMGand Policy: Methodological, David Reneft 39175Conceptual, and Statistical Problems
WPS609 A Sensitivity Analysis of Cross- Ross Levine March 1991 CECMGCountry Growth Regressions David Rnelt 39175